Measurement properties of the German version of the Cambridge examination for mental disorders of older people with Down syndrome and others with intellectual disabilities (CAMDEX-DS)

ABSTRACT

Background

The CAMDEX-DS is an instrument to diagnose Alzheimer’s disease (AD) in Down syndrome consisting of an informant interview and a cognitive test battery (CAMCOG-DS). Measurement properties of the German CAMDEX-DS were investigated.

Method

Fifty-five adults with Down syndrome (19–58 years) participated in this observational study. “Dementia” and “Alzheimer's dementia” (Alzheimer's disease) were diagnosed clinically and operationalised CAMDEX-ICD-10 criteria were applied to evaluate criterion validity. Validity and reliability of the CAMCOG-DS were analysed.

Results

Specificity of the interview was 69–93%; sensitivity 0–80% for “dementia”; and 0-20% for Alzheimer's disease. A complete CAMCOG-DS score was obtained in 85% (item difficulty 0.11–0.96). Construct validity and retest-reliability were low to moderate (τ = .04–.79), inter-rater reliability excellent (τ = .70–.89), internal consistency and selectivity acceptable to excellent.

Conclusions

Currently, the CAMDEX-DS including the CAMCOG-DS are the outcome assessments for assessing dementia in Down syndrome with the best psychometric properties; however, revision is recommended.

KEYWORDS:

• cognition
• dementia
• neuropsychology
• assessment
• Down syndrome
• intellectual disability

Down syndrome is the most common genetic cause for intellectual disability (Mégarbané et al., 2009). People with Down syndrome often have a mild to moderate intellectual disability (Grieco et al., 2015) and a higher risk of developing Alzheimer's disease, compared to the normal population (Lautarescu et al., 2017). Clinical evidence of dementia can be observed in 50–80% of the Down syndrome population above the age of 60 (Hithersay et al., 2017).

In order to diagnose dementia, current cognitive deficits, as well as the deterioration of performance in different cognitive domains have to be documented (Hanney et al., 2009). In English-speaking countries, different tasks and test batteries for diagnosing dementia in people with an intellectual disability have been developed, while only a very limited number of validated instruments is available in German (Loosli et al., 2021; Zeilinger et al., 2022). Assessing dementia symptoms in people with intellectual disabilities can be difficult, as self-reflection, communication, and cooperation are often limited and other confounding factors such as speech and hearing problems or comorbidities that mimic dementia symptoms can impede the diagnostic process.

The CAMDEX-DS (Cambridge Examination for Mental Disorders of Older People with Down syndrome and Others with Intellectual Disabilities) (Ball et al., 2006) is a comprehensive instrument based on a reflective model. It is based on the CAMDEX (Roth et al., 1986) and was adapted for the diagnosis of dementia in people with Down syndrome. It comprises an interview with a caregiver, as well as a neuropsychological examination that includes different tasks suitable for people with mild and moderate intellectual disability (Cambridge Cognitive Examination for Older Adults with Down syndrome, CAMCOG-DS).

The CAMDEX-DS is used in international research projects and its measurement properties have been investigated in the English, Spanish and Portuguese versions (Ball et al., 2004; Esteba-Castillo et al., 2013; Fonseca et al., 2019). A systematic review on informant-based assessment instruments for dementia in people with intellectual disability rated measurement properties of the CAMDEX-DS interview as follows: Sufficient content validity, internal consistency, reliability, criterion validity, construct validity and responsiveness; structural validity and measurement error were indeterminate (Zeilinger et al., 2022). Individual studies showed that the CAMDEX-DS interview has a good inter-rater reliability (Esteba-Castillo et al., 2013) and good concurrent and predictive validity (Esteba-Castillo et al., 2013; Fonseca et al., 2019). Patients with dementia showed decline in CAMCOG-DS scores (Ball et al., 2006; Fonseca et al., 2019). The CAMCOG-DS has a high inter-rater- and test-retest reliability (Esteba-Castillo et al., 2013; Fonseca et al., 2019). The CAMDEX-DS therefore seems to be a valid and reliable instrument to diagnose and monitor dementia in adults with Down syndrome.

Much less is known about measurement properties of the CAMCOG-DS. Given that longitudinal neuropsychological assessments are still the gold standard to objectify decline in Down syndrome, any instrument applied to that end should have good psychometric scores regarding validity and reliability and should be applicable across different levels of cognitive abilities. Since conducting baseline cognitive assessments in young adults with Down syndrome as a reference for future examinations has been recommended (Lautarescu et al., 2017), it is important that the CAMCOG-DS is also valid and reliable in young adults and no ceiling effects appear in this age group. Therefore, we aimed to include not only older, but also younger adults with Down syndrome.

The main rationale for this study is to investigate feasibility and measurement properties of the German CAMDEX-DS. Specifically, we aimed to analyse criterion validity between the CAMDEX-DS interview and clinical diagnosis. Regarding the CAMCOG-DS, we aimed to demonstrate feasibility in younger and older adults with Down syndrome, and to analyse validity and reliability. In terms of construct validity, we hypothesise that there will be a positive relationship between the CAMCOG-DS total score and its sub-scale scores and another cognitive test battery that was developed to assess cognition in patients with severe cognitive deficits (Severe Impairment Battery, SIB, Saxton et al., 1993). Further, we expect that those with cognitive decline will show lower scores in the CAMCOG-DS than those without cognitive decline. Regarding reliability, we investigate internal consistency, split-half reliability, selectivity, test-retest as well as inter-rater reliability between. As many of these measurement properties have not been investigated before, these analyses will expand previous knowledge about the CAMDEX-DS and assessing cognition in people with intellectual disabilities in general. In sum, we expect that the German version will have similar psychometric properties as the versions in other languages.

Taxonomy and reporting are based on the “Consensus-based Standards for the Selection of Health Measurement Instruments (COSMIN)” (Gagnier et al., 2021; Mokkink et al., 2020), the “Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement” (von Elm et al., 2008) and the “Guidelines for Reporting Reliability and Agreement Studies (GRRAS)” (Kottner et al., 2011).

Methods

Study design

An observational, cross-sectional design with an ad-hoc sample of adults with Down syndrome was applied. All patients who visited our outpatient clinic between April 2016 and January 2020 and fulfilled inclusion and exclusion criteria (see below) were included. No estimation of sample size was undertaken before data collection began. A smaller sub-sample had a second CAMCOG-DS assessment after approximately 30 days to investigate test-retest reliability.

Participants

Participants were recruited during consultations at the Ludwig-Maximilians-University Hospital’s outpatient clinic for adults with Down syndrome embedded in the department of neurology. In addition to the study procedures listed below, a clinical routine examination and additional investigations were performed if applicable. Most patients came with neurological, cognitive or behavioural concerns.

Inclusion criteria were a diagnosis of Down syndrome, age ≥18 years and the ability to understand the instructions of the CAMCOG-DS. Further, a caregiver had to be available for the CAMDEX-DS interview. Exclusion criteria were severe structural brain lesions and inability to respond to at least some of the items in the CAMCOG-DS (e.g., due to severe cognitive deficits, no speech production, or lack of cooperation).

The study was conducted in accordance with the Declaration of Helsinki in its latest revision. The study protocol has been approved by the local ethics committee of the Ludwig-Maximilians-Universität München (Project 17-107). Written informed consent was retrieved from all legal guardians and oral informed consent was obtained from patients if possible.

Clinical diagnosis

Two neurologists (GN, OW) independently evaluated all clinical information (medical history, clinical examination, laboratory findings as well as MRI / PET findings if available, except for data from the CAMDEX-DS interview). A diagnosis was reached applying a standardised protocol (Nübling et al., 2022). Possible diagnoses were the ICD-10 diagnoses “dementia”, “mild cognitive impairment” (MCI) as well as cognitive disorders due to secondary causes or a cognitive disorder not otherwise specified (in cases where cognitive decline was equally attributable to two or more diagnoses). Individual diagnoses were compared, and, in cases of discrepancy, cases were reviewed again until a consensus was reached. This consensus was considered the gold standard for further evaluations. In addition, the level of intellectual disabilitsy was rated in accordance with DSM-V criteria by a neuropsychologist (SL).

Instruments

CAMDEX-DS

The CAMDEX-DS consists of two major parts: A structured interview with a caregiver and a neuropsychological test battery, the CAMCOG-DS (Ball et al., 2006). It was translated and culturally adapted according to international standards (Beaton et al., 2000; Epstein et al., 2015), different steps are described in detail in Nübling et al. (2020). To summarise, the English version was translated by two members of our team into German, then a consensus version was developed in our group. It was used with ten patients, adapted if necessary, and then back translated into English by two independent professional translators. Then, two originally English speaking persons compared the back translations and the original English versions, and changes were made in the German version, until the agreement between back translation and original version was defined as satisfying by the consensus group (see Nübling et al., 2020).

CAMDEX-DS interview

The caregiver interview includes 157 questions and assesses the patient’s best level of functioning, the decline of the patient’s cognitive and functional abilities, the current psychological and physical condition and potential confounders (alcohol abuse, medication, etc.). To assess decline, care-givers reported whether the patient has difficulties (e.g., finding words), and whether this was a slight or great decline or no decline (closed questions). A standardised analysis was conducted concerning the presence of dementia syndrome as well as Alzheimer's disease in accordance with ICD-10 as defined in the manual, based solely on the information available from the CAMDEX interview. In brief, a deterioration sufficient to interfere with daily activities had to be reported with a duration of at least six months to confirm a clinical syndrome of dementia. Hereby, memory and at least one other cognitive domain as well as emotional, motivational, or social functioning had to be affected. Delirium had to be ruled out. To confirm a diagnosis of Alzheimer's disease, other possible causes had to be excluded (e.g., psychiatric diseases, stroke, influence of medication).

CAMCOG-DS

The CAMCOG-DS is a standardised test battery and consists of 45 items and 7 subscales. Tasks vary from open questions (e.g., naming current month) to action items (apraxia) or copying tasks (visuo-perception). Tasks were scored according to predetermined categories and examples (typically 0/1/2 points, or 0/1 points). The sub-scales include Orientation (12 points), Language (27 points), Memory (29 points), Attention (9 points), Praxis (18 points), Abstract Thinking (6 points) and Visual Perception (8 points). The maximum score is 109 points.

Severe impairment battery (SIB)

The SIB was developed for the cognitive assessment of people with severe dementia and offers a diverse set of tasks for a low performance level (Saxton et al., 1993). The nine subscales are Attention, Praxis, Orientation, Language, Memory, Orientation towards his/her own name, Social Interaction, Visuo-spatial and Constructive Praxis. A maximum score of 100 can be achieved. It has previously been validated in German (Spiegel et al., 2001) and in persons with Down syndrome (Witts & Elders, 1998).

Procedure

Medical history was collected with the patient and the caregiver before a neurological examination was conducted. Subsequently, the CAMDEX-DS assessments with the patient and the caregiver were conducted separately. The CAMDEX-DS interviews were conducted by neurologists (CP, LF) or by trained student assistants while the patient was examined by a neuropsychologist (SL, EW) with the CAMCOG-DS. Interview and test battery were administered as paper-pencil versions as in the original. After the CAMCOG-DS, the SIB was conducted, if the patient could still concentrate on the tasks and was motivated to continue with the assessments. Two neuropsychologists with >10 years experience conducted the CAMCOG-DS and the SIB. Scoring was discussed on a regular basis to improve reliability.

If a specific consent was given, the conduction of the CAMCOG-DS was recorded with a camera, so that the second neuropsychologist could rate the patient’s performance at a later point in time to assess inter-rater reliability.

Patients were also asked to take part in a second neuropsychological assessment after 30 days, during which the CAMCOG-DS took place in order to calculate test-retest reliability. Patients and care-givers were asked at the second assessment whether they were clinically stable or whether changes occurred from first to second assessment. Patients were assessed in the same room and by the same neuropsychologist at both time points.

Statistical analysis

CAMDEX-DS interview

Interviews were scored according to the ICD-10 diagnostic algorithm defined in the manual (Ball et al., 2006, see above). The result of the interview (criterion for “Alzheimer’s disease” fulfilled yes/no and for “dementia” fulfilled yes/no) was compared to the “gold standard”, the diagnosis established after all clinical examinations were completed. To assess criterion validity, specificity and sensitivity were calculated. In addition, agreement, kappa and Yules Y were calculated to compare the CAMDEX-DS diagnosis with the gold standard. Kappa can be biased by different prevalences of a symptom; therefore, Yules Y should be taken instead of Kappa if it differs from the latter (Wirtz & Kutschmann, 2007).

CAMCOG-DS

Parametric tests were used in the whole sample (N = 47) and non-parametric tests with the small sub-samples (construct validity with SIB, test-retest reliability, inter-rater reliability). P < .05 was used as cut-off for statistical significance. A detailed description of outlier analysis and how feasibility as well as different validity and reliability measures were calculated can be found in Appendix 1. Data were analysed using IBM SPSS Statistics, release 25.0.1.

Results

Participants

In total, 55 individuals with Down syndrome (21 female, 34 male) and their caregivers took part in the study. Age at first consultation ranged from 19 to 58 years (M = 35,5 years, SD = 12,4 years). According to DSM-V criteria, 46% of these participants had a mild intellectual disability, 36% a moderate intellectual disability and 18% a severe intellectual disability. Of the 55 patients included, two had neither an interview nor CAMCOG-DS data. They were excluded from further analyses. From the remaining 53 individuals, six had incomplete CAMCOG-DS data (some items were skipped because of severe cognitive deficits or lacking cooperation). Therefore, the analyses of measurement properties of the CAMCOG-DS included 47 patients and the analyses regarding the CAMDEX-DS interview included 34 participants. Demographic data of these samples can be found in Table 1.

Table 1. Sample demographics.

	Total sample	No cognitive disorder	MCI (ICD-10)	Dementia (ICD-10)	Secondary cognitive disorder	Cognitive disorder nos
Sample
CAMDEX-DS Interview
n	34	17	2	5	5	5
Female n (%)	15 (44%)	10 (59%)	0 (0%)	2 (40%)	1 (20%)	2 (40%)
Age M (SD), (range)	33,5 (11,9), 19–58	26,2 (5,8), 19–41	45,0 (18,4), 32–58	49,8 (4,02), 45–56	26,6 (8,7), 29–50	44,4 (8,7), 29–50
ID (mild/moderate/severe)	17/11/6	15/1/1	0/2/0	0/3/2	2/3/0	0/2/3
CAMCOG-DS
n	47	23	5	7	7	5
Female n (%)	19 (40%)	11 (48%)	2 (40%)	3 (43%)	1 (14%)	2 (40%)
Age M (SD), (range)	35,2 (2,5), 19–58	27,3 (6,1), 19–41	49,8 (10,5), 32–58	52,9 (5,1), 45–58	28,3 (5,0), 19–35	42,2 (9,3), 29–50
ID (mild/moderate/severe)	24/18/5	18/5/0	3/2/0	0/5/2	2/4/1	1/2/2
CAMCOG / SIB
n	20	16	1	0	2	1
Female n (%)	12 (60%)	9 (56%)	1 (100%)	-	1 (50%)	1 (100%)
Age M (SD), (range)	27,0 (7,6), 19–51	26,0 (5,3), 19–38	51	-	22,5 (5,0), 19–26	29
ID (mild/moderate/severe)	16/4/0	14/2/0	1/0/0	-	1/1/0	0/1/0
CAMCOG-DS test-retest
n	22	12	2	3	4	1
Female n (%)	10 (45%)	6 (50%)	1 (50%)	1 (33%)	1 (25%)	1 (100%)
Age M (SD), (range)	33,1 (12,7), 19–58	27,3 (6,7), 20–41	54,5 (4,9), 51–58	53,3 (4,6), 48–56	25,8 (4,8), 19–30	29
ID (mild/moderate/severe)	12/9/1	10/2/0	1/1/0	0/2/1	1/3/0	0/1/0
CAMCOG-DS interrater
n	33	19	3	2	6	3
Female n (%)	15 (46%)	10 (53%)	1 (33%)	1 (50%)	2 (33%)	1 (33%)
Age M (SD), (range)	30,9 (11,2), 19–58	26,4 (6,2), 19–41	49,0 (14,7), 32–58	52,0 (5,7), 48–56	25,8 (4,2), 19–30	38,0 (10,1), 29–49
ID (mild/moderate/severe)	18/15/0	14/5/0	1/2/0	0/2/0	2/4/0	1/2/0

Note. MCI = Mild Cognitive Impairment; nos = not otherwise specified; SIB = Severe Impairment Battery; ID = Intellectual Disability.

CAMDEX-DS interview

A total of 34 CAMDEX-DS interviews were included in the analysis. Caregivers reported changes in memory in half of the patients, and changes in other cognitive functions as well as behavioural and emotional changes in two thirds of patients (see Table 2).

Table 2. CAMDEX-DS informant interview. Fulfillment of ICD-10 criteria for Dementia syndrome and Alzheimer's disease.

	N (%)
Sample total	34 (100%)
Domain G1: Memory impairment	17 (50%)
Domain G2: Decline in other cognitive functions / daily living skills	21 (62%)
Domain G3: Absence of delirium	29 (85%)
Domain G4: Decline in emotional control, motivation or social behavior	22 (65%)
Domain G5: Duration ≥ 6 months	21 (62%)
Dementia syndrome (domains G1-G5 fulfilled)	10 (29%)
Alzheimer’s disease (Dementia syndrome and exclusion of other causes)	3 (9%)

Table 2 shows the number of patients who fulfilled the dementia and Alzheimer's disease criteria according to the ICD-10 criteria defined in the CAMDEX-DS manual. In total, ten patients were classified as having a dementia, but only three patients fulfilled the criteria for Alzheimer's disease.

In this subset of 34 participants, 17 patients received a clinical diagnosis according to the gold standard procedure: Dementia (n = 5), MCI (n = 2), secondary cognitive disorder (n = 5), cognitive disorder not otherwise specified (nos) (n = 5) (see Table 1). An overview of agreement between CAMDEX-DS ICD-10 classification and the clinical diagnosis (gold standard) is displayed in Table 3.

Table 3. Agreement between CAMDEX-DS interview and clinical diagnosis.

		Clinical Diagnosis (Gold Standard)
		ICD-10 no MCI	ICD-10 MCI	Total		ICD-10 no Dementia	ICD-10 Dementia	Total
CAMDEX diagnosis
No Dementia		22	2	24		23	1	24
Dementia		10	0	10		6	4	10
Total		32	2	34		29	5	34
No AD		29	2	31		27	4	31
AD		3	0	3		2	1	3
Total		32	2	34		29	5	34

Note. MCI = Mild Cognitive Impairment; AD = Alzheimer's disease.

Specificity, sensitivity and agreement scores are provided in Table 4. Specificity was in the moderate to high range and did not differ much for the different types of cognitive disorder (specificity between 68% and 79%), so patients with no cognitive disorder were correctly identified in the majority of cases. While sensitivity to detect dementia was high (4/5, 80%), it was not possible to identify MCI (sensitivity of 0%). Altogether, agreement was highest with ICD 10 dementia diagnosis.

Table 4. Specificity, sensitivity and agreements scores of CAMDEX-DS Interview.

	Clinical Diagnosis (Gold Standard)
	Any cognitive decline	MCI	Dementia	Secondary cognitive disorder	Cognitive disorder nos
CAMDEX-DS ICD-10 criteria
Dementia Syndrome
Specificity (%)	88,2	68,8	79,3	72,4	72,4
Sensitivity (%)	47,1	0,0	80,0	40,0	40,0
Kappa	−0,10	−0,11	0,42	0,09	0,09
Yules Y	0,44	n/a*	0,59	0,14	0,14
Agreement	68%	65%	79%	68%	68%
Alzheimer's Disease
Specificity (%)	94,1	90,6	93,1	89,7	93,1
Sensitivity (%)	11,8	0,0	20,0	0,0	20,0
Kappa	−0,02	−0,08	0,16	−0,12	0,16
Yules Y	0,19	n/a*	0,30	n/a*	0,30
Agreement	53%	85%	82%	77%	82%

Note. N = 47. MCI = Mild Cognitive Impairment; NCD = neurocognitive disorder; nos = not otherwise specified. *Could not be calculated as one value was 0.

When applying the CAMDEX-DS ICD-10 criteria for Alzheimer's disease, specificity considerably increased (90-93% for all types of cognitive disorder, 94% for identifying any cognitive disorder). However, sensitivity to diagnose dementia notably decreased (20%).

CAMCOG-DS

Feasibility

A complete CAMCOG-DS total score could be obtained from 47 of the 55 included patients (85,4%). Of the eight patients of whom no full score could be obtained, five had a severe intellectual disability and two a moderate intellectual disability. One had a dementia and three a psychiatric disorder. Six were males, and three were older than 45 years. Altogether, the CAMCOG-DS is tolerated well in patients with mild and moderate intellectual disability.

Results regarding item difficulty are displayed in Table A1 in Appendix 2. Very easy items were 172 (Orientation to name, 96% correct responses), 182 (Telling the name of two objects, 93% correct) and 200 / 201 (Drawing a circle / square, 89% correct in each item). Very difficult items could be found in the Memory scale: Item 191 (telling the name of “the Beatles”, 11% correct responses), 194 (telling the name of the German Federal President, 13% correct) and 192 (telling the name of Lady Diana, 16% correct).

Descriptives and individual differences

The 47 patients included in the CAMCOG-DS analyses were between 19 and 58 years. The CAMCOG-DS total score ranged between 11 and 88. Younger participants performed significantly better than older ones, and those with mild intellectual disability performed significantly better than those with moderate or severe intellectual disability (see Table 5).

Table 5. Descriptives and differences between age groups and intellectual disability (ID) groups.

	Total Sample				Age ≤45			Age > 45			t-Test		Mild ID			Moderate ID			Severe ID			ANOVA
	N = 47				N = 34			N = 13					N = 24			N = 5			N = 18
Measure	M	SD	Range		M	SD		M	SD		t(45)		M	SD		M	SD		M	SD		F(2,44)	Post-hoc (Bonferroni)
CAMCOG-DS
Orientation	7,60	3,37	1–12		8,21	3,37		6,00	2,92		2,077*		9,17	2,99		6,67	2,97		3,40	1,14		9,93***	Mild > (Mod = Sev)
Language	17,55	5,19	6–27		18,56	4,96		14,92	5,02		2,242*		20,41	4,26		16,17	3,22		8,80	2,68		21,70***	Mild > Mod > Sev
Memory	10,13	5,58	0–22		11,44	5,23		6,69	5,15		2,796**		13,71	4,56		7,39	3,68		2,80	1,79		21,67***	Mild > (Mod = Sev)
Attention	6,70	2,36	0–9		7,03	2,08		5,85	2,88		1,352^+		7,83	0,87		6,22	2,78		3,00	1,00		14,92***	Mild > Mod > Sev
Praxis	11,30	3,92	0–17		12,06	3,34		9,31	4,72		1,927*		13,21	2,45		10,78	3,21		4,00	2,92		22,74***	Mild > Mod > Sev
Abstract Thinking	2,09	1,92	0–6		2,15	1,71		1,92	2,47		0,301		2,63	2,14		1,94	1,43		0,00	0,00		4,55*	Mild > Sev; Mod = Sev
Perception	3,91	1,68	0–7		4,24	1,62		3,08	1,61		2,358*		4,71	1,37		3,50	1,47		1,60	1,14		11,99***	Mild > Mod > Sev
Total Score	59,28	20,10	11–88		63,68	18,44		47,77	20,37		2,571**		71,67	13,31		52,67	14,48		23,60	5,18		30,77***	Mild > Mod > Sev

Note. *** p < .001, ** p < .01, * p < .05, + p < .10. One-tailed t-tests, as higher performance in younger patients could be expected. N = 47.

Of this subsample, 51% received a clinical diagnosis (see Table 1). Significant age differences were revealed between diagnostic groups: Those with dementia were oldest, while those with a secondary or no cognitive disorder were youngest (ANOVA, F(4,42) = 28,978, p < .001, η² = .734).

Regarding the CAMCOG-DS total score, those without a cognitive disorder performed better than those who had any cognitive disorder (t(37,295) = 4,687, p < .001, d = 1,360). Regarding different diagnostic subgroups, those without a cognitive disorder generally performed best and those with a dementia performed worst (see Table 6). The ANOVA revealed significant between-group differences (F(4,42) ≥ 4,416, p < .01, η² ≥ .296) regarding the CAMCOG-DS total score and all subscales, except the attention subscale, for which only marginal significant differences were found (F = 2,327, p = .072, η² = .180). Bonferroni post-hoc analyses showed that regarding the CAMCOG-DS total score, patients without cognitive disorders performed significantly better than those with a dementia, a secondary cognitive disorder, or those with an unclear/unknown cognitive disorder (p < .05), but did not differ from those with MCI (p = .100). MCI patients did not differ from those with dementia (p = .184).

Table 6. Descriptives of different diagnostic groups.

	No cognitive disorder			Any cognitive decline			MCI			Dementia			Secondary cognitive disorder			Cognitive disorder nos
	N = 23			N = 24			N = 5			N = 7			N = 7			N = 5
Measure	M	SD		M	SD		M	SD		M	SD		M	SD		M	SD
CAMCOG-DS
Orientation	9,52	2,68		5,80	2,94		8,80	2,78		4,57	0,98		4,86	3,44		5,60	2,7
Language	20,26	3,93		14,96	4,96		18,40	2,88		13,43	5,09		13,71	5,91		15,40	4,51
Memory	13,04	4,30		7,33	5,29		9,20	6,83		5,57	3,41		7,29	5,28		8,00	6,67
Attention	7,61	1,03		5,83	2,91		7,00	3,39		5,57	2,70		5,71	3,20		5,20	2,95
Praxis	13,22	2,32		9,46	4,28		12,00	2,45		9,00	4,58		9,14	4,67		8,00	4,9
Abstract Thinking	2,30	1,66		1,88	2,15		4,80	1,30		0,86	1,07		1,00	1,29		1,60	2,61
Perception	4,87	1,32		3,00	1,47		4,40	0,89		2,57	1,62		2,57	1,62		2,80	0,84
Total Score	70,83	11,84		48,21	20,31		64,60	15,69		41,57	15,25		44,29	23,16		46,60	23,01

Note. N = 47. MCI = mild cognitive impairment; nos = not otherwise specified.

Similarly, those without a cognitive disorder performed better in six of the seven subscales compared to those with any cognitive disorder (t(45) ≥ 2,806, p < .01, d ≥ 0,823). No differences appeared regarding Abstract Thinking (t(45) = 0,763, p = .450, d = 0,203). Healthy patients performed better than those with dementia regarding Orientation, Language, Memory and Visual Perception (p < .05) and marginally better regarding Praxia (p < .100). Those with MCI performed better in Abstract Thinking (p < .05) than those without a cognitive disorder. Apart from this, there were no differences between those without a cognitive disorder and MCI patients. Patients with an MCI performed better in Abstract Thinking (p < .01) and marginally better in Orientation (p < .100) than patients with a dementia.

As age differences were found in CAMCOG-DS performance, age was included as a covariate in an ANCOVA. Group differences mainly remained significant (F(4,41) ≥ 2,916, p < .05, ηp2 ≥ 0,221). Only regarding the attention subscale, no significant differences could be revealed (F(4,41) = 1,846, p = .139, ηp2 = 0,153).

Construct validity

The 20 patients who completed the SIB were younger than the total sample and had milder forms of intellectual disability (see Table 1). SIB scores were high and a ceiling effect was evident in some of the scales (see Table 7). In line with the hypothesis, non-parametric analyses revealed that the relationship between CAMCOG-DS and SIB total scores was moderate (τ = .50, 95% CI [.27, 1.00], p < .01). Regarding subscales, relationships were moderate for Orientation (τ = .45, 95% CI [.21, 1.00], p < .05), Language (τ = .47, 95% CI [.23, 1.00] p < .01) and Praxis (τ = .42, 95% CI [.14, 1.00], p < .05) and low for Attention (τ = .30, 95% CI [.04, 1.00], p = .147). There was no relationship regarding Memory scales (τ = .08, 95% CI [-.19, 1.00], p = .652). The CAMCOG-DS Visual Perception score was not related to the SIB Visuospatial scale (τ = .04, 95% CI [-.21, 1.00], p = .881). As expected, participants with cognitive decline showed lower CAMCOG-DS scores than those without cognitive decline. Those with dementia showed the lowest scores (see results in Section “Descriptives and Individual Differences’ as well as Table 6).

Table 7. Descriptives and group characteristics SIB.

Measure	M	SD	Range
Social Interaction	6,00	0,00	6–6
Memory	12,10	2,17	8–14
Orientation	5,45	0,95	3–6
Language	42,70	3,21	36–46
Attention	5,30	0,92	3–6
Praxis	7,05	1,82	1–8
Visuospatial Ability	7,60	0,75	6–8
Construction	3,80	0,62	2–4
Orientation to Name	1,80	0,52	0–2
Total Score	91,80	7,21	73–99

Note. N = 20.

Interrater-reliability

Results of interrater-analyses of the 33 participants are provided in Table A1 in Appendix 2 (for sample characteristics, see Table 1). Non-parametric correlations between subscale and total scale scores of both raters were between τ = .70 (95% CI [.55, .81]) for Praxis and τ = .89 (95% CI [.83, .93]) for Attention and show an excellent agreement between the ratings. On the item level, both non-parametric correlations as well as percentage as measures of agreement showed mostly good to excellent interrater agreements. Of the 46 items, 22 had correlations ≥ .80 and in 34 items, agreement was ≥ 80%. In contrast to the other items, rather low agreements were obtained in the Praxis scale (202: drawing a house; 207: showing how to use scissors).

Test-retest reliability

In total, 22 patients were retested with the CAMCOG-DS after 4–5 weeks (26–42 days, M = 32,8 days, SD = 4,4), see Table 1 for sample description. Means for Total Score, Language and Memory were significantly higher at the second assessment (M_Test1 = 59,73, SD_Test1 = 18,45, M_Test2 = 65,59, SD_Test2 = 18,92, Z ≤ −2,483, p < .05). Also, all subscale score means were higher – at least on a descriptive level – at the second assessment. All but two patients had a higher total score at the second assessment. Results regarding test-retest reliability are depicted in Table A1 in Appendix 2. Non-parametric analyses revealed that scores were adequate for the total score (τ = .78, 95% CI [.65, 1.00], p < .001) and for the subscales Orientation (τ = .79, 95% CI [.67, 1.00], p < .001), Language (τ = .75, 95% CI [.61, 1.00], p < .05) and Memory (τ = .75, 95% CI [.61, 1.00] p < .001). The score of the Visual Perception scale was in the marginal range (τ = .65, 95% CI [.48, 1.00], p < .001). Only low scores could be obtained regarding Attention (τ = .56, 95% CI [.36, 1.00] p < .05), Praxis (τ = .58, 95% CI, [.38, 1.00], p < .01), and Abstract Thinking (τ = .43, 95% CI [.20, 1.00], p < .05). Highest values were reached in items of the Orientation scale (see Table A1, Appendix 2). Only 8 of 46 items (17,4%) had a score above 0.7.

Split-half-reliability and internal consistency

Item Selectivity was high (> .50) for most items, which is in line with the assumption of unidimensionality. An excellent split-half reliability was revealed in the correlation analysis (r = .96, 95% CI [.87, .96] p < .001). Cronbach’s Alpha was excellent (.94, 95% CI [.91, .96]) for the CAMCOG-DS total score, and good for the Language (.83 95% CI [.75, .90]) and Praxis (.80, 95% CI [.71, .88]) subscales. Orientation (.78, 95% CI [.67, .86]), Memory (.74, 95% CI [.61, .84]), Attention (.70, 95% CI [.52, .83]), and Abstract thinking (.75, 95% CI [.59, .85]) had also acceptable values, but the score for Visual Perception was low (.47, 95% CI [.04, .70]).

Discussion

This investigation of measurement properties of the CAMDEX-DS aimed at addressing the lack of validated diagnostic tools for dementia in Down syndrome in the German language. To date, only the Dementia Test for Individuals with Intellectual Disabilities (DTIM, Müller & Kuske, 2020) is available in German, which combines the informant questionnaire Dementia Screening Questionnaire for Individuals with Intellectual Disabilities (Deb et al., 2007) and a newly developed neuropsychological test battery. Regarding the latter, preliminary data are available regarding psychometric criteria (Kuske et al., 2017). With the emergence of disease-modifying treatment strategies, there will be a need for reliable diagnostic tools for dementia in Down syndrome in the foreseeable future.

Informant interview of the CAMDEX-DS

The 157 questions of the CAMDEX-DS informant interview cover a wide spectrum of symptoms not only related to Alzheimer's disease, but also to important differential diagnoses such as delirium, psychiatric disorders and other medical conditions that might mimic Alzheimer's disease. While the manual provides clear instructions as to which questions are to be analysed concerning the different domains of dementia diagnosis as well as differential diagnoses, it provides no cut-off values. Therefore, to minimise the risk for false-positive diagnoses, any answer pointing towards a differential diagnosis was deemed sufficient to preclude a diagnosis of dementia or Alzheimer's disease in this study.

This approach decreased the sensitivity for “dementia” to 80% since one patient with a clinical diagnosis of Alzheimer's disease had a positive answer in one question related to delirium, while delirium could clearly be excluded in the clinical interview and neurological examination. Moreover, the strict application of these criteria decreased the sensitivity for “Alzheimer's disease” to 20%. Of the four false-negative patients, two had a concomitant diagnosis of depression, which was clinically deemed not relevant to the observed cognitive decline, one had a (non-confirmed) suspicion of neurosyphilis, and the aforementioned patient with a suspicion of delirium had no other exclusion criteria.

In contrast to sensitivity, specificity was at 79% when applying the “dementia” criteria, and at 93% when applying the Alzheimer's disease criteria, which is comparable to previous validation studies of this instrument (Ball et al., 2004; Fonseca et al., 2019). Therefore, patients without dementia were correctly classified in most cases. In sum, criterion validity with the gold standard was moderate to high, while sensitivity was moderate for dementia and low for Alzheimer's disease.

Recently, threshold scores for Part 2 of the interview, cognitive and functional decline, have been published (Beresford-Webb et al., 2021). They differentiate between healthy persons with Down syndrome and those with prodromal Alzheimer's disease or Alzheimer's disease, respectively, and might improve diagnostic accuracy of this instrument.

Neuropsychological test battery CAMCOG-DS

Another main aim of this study was to evaluate the psychometric criteria of the German CAMCOG-DS. Applicability and feasibility of the CAMCOG-DS were good (85% completion rate). Very difficult items were scarce and possibly related to inadequate cultural adaption (recalling the names of Lady Diana and the Beatles). A good mixture between easy and difficult items is good to keep motivation and cooperation on a high level and shows that the CAMCOG-DS can be used in people with a wide range of cognitive abilities, and nonetheless has only few items with ceiling or floor effects.

Evaluating construct validity with a validated neuropsychological test battery was challenging due to the lack of validated German tools. The DTIM (Müller & Kuske, 2020), which is comparable to the CAMCOG-DS in certain aspects, was not published yet when data collection for the present study started. The expected positive relationship between CAMCOG-DS and SIB was confirmed regarding total score, Orientation and Praxis, but not regarding the other sub-scales. Construct validity is therefore in the low to moderate range. However, patients showed ceiling effects in the SIB; therefore, it might have been too easy for some of the patients. The SIB had previously shown ceiling effects in a Down syndrome cohort similar to ours (Witts & Elders, 1998). Therefore, the relationship between SIB and CAMCOG-DS should probably be re-evaluated in an older and cognitively less able sample, and also construct validity should be evaluated in a older sample.

Further, as in previous studies with the CAMCOG-DS or its precursors, better performance in younger than in older participants (Hon et al., 1999) and in those with mild compared to moderate or severe intellectual disability (Esteba-Castillo et al., 2013) was observed. Additionally, patients without a cognitive disorder performed better than those of other diagnostic groups, especially better than those with dementia, which is in line with a previous study (Benejam et al., 2020). Although based on small sub-groups, these expected results regarding individual differences confirm our hypotheses and underline the construct validity of the CAMCOG-DS, as better performance in healthy, younger and cognitively more able patients can be expected in most of the CAMCOG-DS tasks.

Overall, reliability scores were mostly adequate to excellent. Interrater-reliability coefficients were mostly good to excellent and comparable with other studies (Esteba-Castillo et al., 2013; Fonseca et al., 2019). It was lower when scoring criteria were only vaguely formulated (e.g., ideomotor praxis, definition of “bridge”). Test-retest reliability was adequate. Most patients showed small retest effects after a few weeks, which can be due to a learning effect, but also due to a familiarisation to the clinic environment and neuropsychologist. Retest effects have previously been noted in adults with intellectual disabilities (Kuske et al., 2017). Internal consistency and item selectivity were acceptable to excellent and similar to a previous validation study (Esteba-Castillo et al., 2013), while split-half reliability was excellent.

Practical relevance and generalisability

The CAMDEX-DS is the only instrument to diagnose dementia in people with intellectual disabilities that includes an informant interview and a cognitive test-battery and has been thoroughly investigated regarding psychometric properties in German as well as in other languages. Therefore, it is suited for longitudinal clinical assessments, but also can be used in international studies. However, training and clinical experience of users is essential to receive valid and reliable scores. We assume that the results of this study are generalisable to other patients with intellectual disabilities.

Limitations and future directions

Interpretation of results might be limited by the rather small sample size and a possible overrepresentation of younger patients and those with mild intellectual disability. Therefore, there might be an increased risk for false positive or negative findings. However, our patient collective covers a broad range of age and degree of intellectual disability as well as healthy and diseased patients, and thus represents the patient collective seen in neurologic and psychiatric outpatient departments.

While previous studies regarding the CAMDEX-DS have already evaluated the properties of the informant interview, the strength of this study lies in the focus on the measurement properties of the neuropsychological test battery.

Future studies of measurement properties of instruments for people with intellectual disabilities should include larger samples. Revisions of the CAMCOG-DS should include updated test material and clearer instructions for scoring to improve reliability.

Finally, the study was designed without consideration of the COSMIN framework and therefore, some of the proposed aspects could not be adapted.

Conclusion

At the moment, the CAMDEX-DS including the CAMCOG-DS are the outcome assessments for assessing dementia in Down syndrome with the best psychometric properties; however, revision is recommended. The informant interview provides comprehensive disease-related information, but must be interpreted in conjunction with a clinical assessment.

Acknowledgements

We thank Nina Smrzka for help with study coordination and the patients and their caregivers for taking part in this study.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

Data are available from the corresponding authors upon reasonable request.

Additional information

Funding

This work was supported by the Fondation Jérôme LeJeune; the VERUM Foundation; the Else Kröner-Fresenius Stiftung [2020_EKEA.09]; the Deutsche Forschungsgemeinschaft DFG, German Research Foundation, under Germany’s Excellence Strategy within the framework of the Munich Cluster for Systems Neurology [EXC 2145 SyNergy – ID 390857198]; and the MOMENTE program for excellent young scientists of the medical faculty of the LMU.

Appendices

Appendix 1

Statistical analysis

Outlier analysis

Extreme values in the data were checked for by transforming the CAMCOG-DS and SIB total and subscale scores into z-scores in the respective samples. All total scores were within z +/- 3 SD from the respective sample mean. Regarding subscales, four scores were between −3 SD and −4.01 SD. However, no data were excluded, as this is a validation study and explicitly should include a broad range of cognitive ability levels.

CAMCOG-DS

To investigate feasibility, completion rate as well as item difficulty was analysed. Item difficulty was calculated according to Fisseni (1997): For dichotomous items, the difficulty index is the proportion of participants from all participants who solved the item correctly (p = N_C / N). For items with more than two possible scores (i.e., recall of pictures, score range from 0-6), the following formula was used: p_m = (∑x) / (∑x_max). Further, differences in cognitive performance between age groups and regarding level of intellectual disability were analysed using t-tests and one-factorial analyses of variance (ANOVA) in the total sample. To analyse differences between diagnostic groups, ANOVAs as well as one-factorial analyses of covariance (ANCOVA) were used.

Construct validity was assessed via hypothesis testing. We expected a significant, positive, at least moderate relationship between the CAMCOG-DS and the SIB (.40 – .60). The relationship was analysed with Kendall’s tau-b (τ), a non-parametric measure, as the subsample was small (n = 20) and data did not meet normality assumptions. Total scores and subscales were compared, if applicable and rated as proposed in Fisseni (1997). As the SIB has two different subscales that can be subsumed under the term “construction/praxis”, they were summed up and correlated with the CAMCOG-DS praxis subscale. Further, the hypothesis regarding diagnostic group differences was tested with t-Tests, ANOVAs and ANCOVAs.

For inter-rater reliability, scores of original and video ratings were analysed using Kendall’s tau-b (τ) and percentage scores (McHugh, 2012; Strauss et al., 2006). Of 41, eight videos were incomplete: Four because of lack of cooperation, two assessments were interrupted and the recording was stopped, and two videos were incomplete due to technical reasons. Therefore, 33 full videos could be evaluated. There were 14 occasions (in 11 patients) during which something was not understandable (8x in four different items) or visible (4x in one item) or the item was not shown in the video (2x in two different items) or not asked (1x in one item) and therefore, not rated. In these 16 cases, the scores were coded as missing.

Test-Retest-Reliability was calculated for the total scores, subscale scores and for all items separately. As the subsample was small (n = 22) and score distributions did not meet normality assumptions, the non-parametric measure Kendall’s tau-b (τ) was used in the correlation analyses.

To analyse internal consistency, first, split-half reliability was calculated. The 46 CAMCOG-DS items were split into two halves to account for effects of practice and fatigue and to equally represent all cognitive domains. Split-half reliability was calculated via Pearson correlation between the total scores and then corrected with the Spearman-Brown-Formula (Fisseni, 1997). Further, Cronbach’s α was calculated for each scale and the CAMCOG-total score. Item selectivity was calculated with corrected item-scale correlations for each item within the respective scale and with the CAMCOG-total score. Items were rated according Fisseni (1997).

Reliability coefficients were rated according to recommended guidelines (Strauss et al., 2006; internal reliability: Cicchetti, 1994; Cicchetti & Sparrow, 1990; inter-rater reliability: Cicchetti & Sparrow, 1981). Kappa and Yules Y coefficients were also rated as previously recommended (Landis & Koch, 1977; Wirtz & Kutschmann, 2007).

Appendix 2

Table A1. Item analysis of CAMCOG-DS.

Scale	Item	Description	Item Difficulty	Retest Reliability	Interrater Reliability		Item Selectivity
			p	Kendall's Tau-b	Kendall's Tau-b	Percentage	Corrected item- scale correlation	Corrected item-test correlation
Orientation
	172	Name	0.96	0.81	0.49	0.94	0.30	0.46
	173	Day	0.76	0.39	0.87	0.94	0.49	0.56
	174	Month	0.53	0.81	0.91	0.94	0.60	0.61
	175	Year	0.52	0.87	0.86	0.91	0.71	0.69
	176	Address	0.34	0.71	0.67	0.73	0.50	0.51
	177	Town	0.69	0.33	0.87	0.91	0.58	0.62
Language
	178	Nod	0.85	0.23	0.80	0.94	0.33	0.37
	179	Ear	0.59	0.34	0.82	0.85	0.61	0.57
	180	Ceiling	0.77	0.43	0.82	0.91	0.65	0.64
	181	Shoulder	0.34	0.55	0.65	0.82	0.53	0.41
	198	Eyes	0.64	0.62	0.48	0.76	0.52	0.58
	199	Hand	0.64	0.70	0.61	0.82	0.23	0.31
	182	Objects	0.93	−0.05	0.23	0.67	0.29	0.25
	183	Pictures	0.87	0.64	0.83	0.82	0.53	0.61
	184	Fluency	0.52	0.63	0.87	0.88	0.60	0.61
	185	Hammer	0.55	0.45	0.61	0.82	0.54	0.48
	186	Chemist	0.70	0.79	0.90	0.97	0.45	0.51
	187	Bridge	0.29	0.59	0.58	0.70	0.46	0.40
	188	Repetition	0.55	0.51	0.76	0.82	0.56	0.68
Memory
	189	Recall pictures	0.25	0.18	0.92	0.88	0.68	0.66
	190	Recognition	0.63	0.48	0.86	0.70	0.55	0.66
	204	Register name	0.49	0.34	0.70	0.79	0.23	0.31
	205	Register address	0.30	0.58	0.87	0.94	0.37	0.41
	211	Recall name	0.24	0.43	0.81	0.85	0.42	0.47
	212	Recall address	0.25	0.61	0.87	0.91	0.62	0.58
	191	Lennon	0.11	0.57	0.72	0.91	0.32	0.31
	192	Diana	0.16	0.69	0.85	0.97	0.36	0.34
	193	German Chancellor	0.62	0.82	0.79	0.82	0.41	0.44
	194	German Federal President	0.13	0.84	0.50	0.88	0.30	0.21
Attention
	195	Twenty	0.80	0.84	0.93	0.97	0.62	0.70
	196	Fingers	0.86	0.54	0.71	0.91	0.68	0.56
	197	Digit-span	0.59	0.37	0.88	0.91	0.60	0.53
Praxis
	200	Circle	0.89	n/a^+	−0.03	0.94	0.62	0.47
	201	Square	0.89	0.69		1.00	0.78	0.67
	202	House	0.48	0.53	0.66	0.19	0.33	0.49
	203	Clock	0.44	0.60	0.69	0.72	0.51	0.49
	206	Wave	0.85	0.46	0.62	0.91	0.56	0.60
	207	Scissors	0.61	0.54	0.04	0.36	0.33	0.32
	208	Toothbrush	0.70	0.07	0.61	0.76	0.66	0.64
	209	Paper	0.59	0.47	0.71	0.73	0.59	0.67
	210	Envelope	0.76	0.59	0.85	0.94	0.67	0.66
Abstract Thinking
	213	Fruit	0.45	0.40	0.86	0.91	0.56	0.43
	214	Clothing	0.37	0.39	0.83	0.84	0.67	0.52
	215	Furniture	0.22	0.49	0.72	0.88	0.51	0.40
Perception
	216	People	0.63	0.60	0.90	0.94	0.35	0.68
	217	Unusual views	0.44	0.66	0.72	0.67	0.35	0.53

Note. N = 47. ⁺Index could not be calculated, as one parameter was constant.